|Numéro de publication||US7219157 B2|
|Type de publication||Octroi|
|Numéro de demande||US 09/815,942|
|Date de publication||15 mai 2007|
|Date de dépôt||23 mars 2001|
|Date de priorité||23 mars 2001|
|État de paiement des frais||Payé|
|Autre référence de publication||US20020138664|
|Numéro de publication||09815942, 815942, US 7219157 B2, US 7219157B2, US-B2-7219157, US7219157 B2, US7219157B2|
|Inventeurs||Stephen M. Blott, Jose′ Carlos Brustoloni, Clifford E. Martin|
|Cessionnaire d'origine||Lucent Technologies Inc.|
|Exporter la citation||BiBTeX, EndNote, RefMan|
|Citations de brevets (6), Citations hors brevets (7), Référencé par (18), Classifications (19), Événements juridiques (3)|
|Liens externes: USPTO, Cession USPTO, Espacenet|
1. Technical Field of the Invention
This invention generally relates to operating systems and, more particularly, to an apparatus and method for providing an application programming interface (API) for network applications.
2. Discussion of the Background Art
Broadly, Application Programming Interfaces (APIs) are those language and messaging formats that define how applications interact with an operating system, and with functions in other applications, communication systems, and hardware drivers. Network applications are those applications that, unlike host applications, process packets whose source and destination are network nodes different from that where the application runs. Network programming is the development of network applications.
Existing networks typically lack platforms that are well suited for network programming. Network hosts usually support only host applications. On the other hand, network bridges, switches, and routers usually execute fixed or configurable network applications and do not provide an API that would enable users to develop new applications.
Without network programming platforms, it can be difficult to prototype or realistically test many desirable new protocols or services. Several protocol enhancements can be demonstrated by analysis or simulation before, e.g., manufacturers implement them and include them among router configuration options. However, certain new services, such as billing (charging users for their network usage), may not be similarly amenable to analysis or simulation. Because the adequacy of a billing strategy depends on how users react to it, and such reaction is difficult to predict or model before observation, simulations may not be convincing. To test such new services, actual deployment and observation in the field are necessary.
Within the context of a network programming API implemented on a personal computer (PC) platform, several features are deemed to be desirable; namely (1) support for configuring the platform as a bridge, router or host; (2) general purpose functionality; (3) support for user-level applications; (4) the ability to pass packets between applications and the operating system without copying; and (5) reducing the number of system calls and interrupts generated per packet.
Unfortunately, existing PC operating systems do not provide APIs that meet all the above requirements. These and other disadvantages associated with the prior art are described in more detail in a paper authored by Blott et al. entitled “NetTap: An Efficient and Reliable PC-Based Platform for Network Programming” published on Mar. 26, 2000 in the Institute of Electrical and Electronic Engineers (IEEE) Proceedings of OPENARCH '2000.
An application programming interface (API) is described that allows creating or destroying one or more pairs of data structures for asynchronously passing between the operating system and a network application pointers to packet buffers mapped to both parties. A network application may use such a pair to: (1) input packets received from a specified network interface, and output packets to be processed by the operating system's network layer as received packets; (2) input packets output by the operating system's network layer to a specified network interface, and output packets to be actually sent by that interface; (3) input packets received and output packets to be sent by a specified network interface; or (4) input allocated and output deallocated packet buffers. The operating system automatically reclaims buffers allocated to a network application when the latter terminates. The API also allows a network application to sleep until the operating system inserts or removes a packet buffer pointer into or from one said data structure. The API may be used in systems configured as host, bridge, switch, or router.
An API according to an embodiment of the invention comprises first and second data structures associated with a network interface in communication with a network, the first and second data structures being mapped to an operating system and a network application, wherein: packets to be passed from the operating system to the network application are stored in a buffer and referenced via respective pointers within the first data structure, the first data structure pointers being inserted into the first data structure by the operating system prior to network layer processing, the first data structure pointers being removed by the network application, insertion and removal of the first data structure pointers being asynchronous with respect to each other; and packets to be processed as received packets by the network layer of the operating system are stored in a buffer and referenced via respective pointers within the second data structure, the second data structure pointers being inserted into the second data structure by the network application, the second data structure pointers being removed by said operating system, insertion and removal of the second data structure Dointers being asynchronous with respect to each other.
An API according to another embodiment of the invention for network applications, which applications can process packets whose source and destination nodes are nodes different from that where the application runs, the API comprising a primitive for creating a first and a second data structures associated with a specified network interface, if the data structures do not exist, and mapping the data structures both to the operating system and a specified network application, wherein the specified network interface receives and sends packets from and to a network, each packet is stored in a buffer mapped both to the operating system and the specified network application, the operating system inserts into and the specified network application may remove from the first data structure a pointer to each buffer containing a packet that the operating system's network layer outputs to the specified network interface, before the network interface sends the packets, the insertions and removals being asynchronous with respect to each other, and the specified network application may insert into and the operating system removes from the second data structure a pointer to each buffer containing a packet that the specified network interface sends to the network, the insertions and removals being asynchronous with respect to each other.
An API according to an embodiment of the invention for network applications, which applications can process packets whose source and destination nodes are nodes different from that where the application runs, the API comprising a primitive for creating a first and a second data structures associated with a specified network interface, if the data structures do not exist, and mapping the data structures both to the operating system and a specified network application, wherein the specified network interface receives and sends packets from and to a network and does not require a coprocessor, the specified network application requires supervisor privileges, every packet is stored in a buffer mapped both to the operating system and every network application, the operating system's network and higher protocol layers do not process any packets that the specified network interface receives or sends, the operating system inserts into and the specified network application may remove from the first data structure a pointer to each buffer containing a packet that the specified network interface receives from the network, the insertions and removals being asynchronous with respect to each other, and the specified network application may insert into and the operating system removes from the second data structure a pointer to each buffer containing a packet that the specified network interface sends to the network, the insertions and removals being asynchronous with respect to each other.
The teachings of the present invention can be readily understood by considering the following detailed description in conjunction with the accompanying drawings, in which:
To facilitate understanding, identical reference numerals have been used, wherever possible, to designate identical elements that are common to the figures.
An illustrative embodiment (herein referred to as “NetTap”) of a network programming API suitable for efficiently prototyping, field-testing, and deploying new network protocols and services is disclosed below. NetTap was implemented by the inventors on low-cost PC (Personal Computer) hardware and the freely available FreeBSD operating system. See, e.g., G. Lehey, The Complete FreeBSD, Walnut Creek CD ROM Books (2d ed. 1997), all relevant portions of which are herein incorporated by reference.
Like FreeBSD's network programming APIs, the NetTap API can be used by a system administrator. Unlike FreeBSD's APIs, however, the NetTap API supports the desirable features enumerated above for network programming APIs for PCs.
Various discussions associated with the prior art and information related to the present invention may be found in a paper by Blott et al. entitled “NetTap: An Efficient and Reliable PC-Based Platform for Network Programming” published on Mar. 26, 2000 in the Institute of Electrical and Electronic Engineers (IEEE) Proceedings of OPENARCH '2000, which paper is incorporated herein by reference in its entirety.
A circular queue comprises a first data structure and a second data structure. Specifically, first and second data structures comprise a detour and revert queue respectively which are mapped to both the operating system 102 and network application 108. More specifically, an input tap 118 comprises a detour queue 104A and a revert queue 106A. Detour queue 104A is coupled to the network interface 114 and network application 108. Revert queue 106A is coupled to network application 108 and TCP/IP implementation 112.
An output tap 120 comprises a detour queue 104B and a revert queue 106B. Detour queue 104B is coupled to TCP/IP implementation 112 and network application 108. Revert queue 106B is coupled to network application 108 and to the network interface 114.
It will be appreciated by those in the art that the invention can be practiced with a plurality of revert and detour queues. In addition the invention can also be practiced with a plurality of network interfaces and network applications.
The above mentioned queues can contain packet pointers 410 (See
It should be noted that network applications are those applications that, unlike host applications, process packets whose source and destination nodes are different from those where the application runs.
As incoming packets arrive, the packets are stored in memory. Data structures are created wherein pointers are placed in the data structures by the operating system 102. Specifically, pointers that point to the location of the incoming packets in memory are placed in detour queue 104A and taken out of detour queue 104A by the network application 108. After the network application 108 has processed a packet, the network application 108 can insert a pointer 410 to the packet buffer 402 in one of three queues. First, the network application 108 can insert the pointer 410 in revert queue 106A. In this case, the operating system's TCP/IP implementation processes the packet as received from the network 116. Second, the network application 108 can insert the pointer 410 in revert queue 106B. In this case, the network interface 114 sends the packet to the network 116. Third, the network application 108 can insert the pointer 410 in the mbuf_dealloc queue 304. In this case, the operating system deallocates the packet buffer 402.
The main characteristics of the NetTap API are: (i) all packet buffers are mapped both to the system and to network applications; (ii) instead of passing copies of packets to one another, system and network applications exchange pointers to packets; and (iii) system and network applications communicate asynchronously via a number of circular queues, thus avoiding system call overheads in normal cases. The system accesses queues asynchronously when a network interface interrupt occurs or returns, a NetTap system call returns, the system preempts or puts to sleep a NetTap application, or the system is idle. Because each queue contains only pointers that are enqueued by the system and dequeued by a network application, or vice-versa, there is no risk of conflict between system and application accesses. Therefore, system accesses do not have to be synchronized with respect to application accesses.
In systems derived from BSD Unix (see M. McKusick, et al., the Design and Implementation of the 4.4 BSD Operating System, Addison-Wesley (1996), all relevant portions of which are herein incorporated by reference), including FreeBSD, network interfaces and protocols hold packets in buffers known as “mbufs”, as illustrated in
As previously discussed above with reference to
NetTap network applications 108 allocate mbufs by dequeueing the mbufs' pointers 410 from mbuf_alloc 302 using the dequeue routine 206, and deallocate mbufs by enqueueing the mbufs' pointers 410 in mbuf_dealloc 304 using the enqueue routine 204. The mbuf_map primitive 208 includes an argument, “mbuf_prealloc” (not shown), that specifies the minimum number of mbuf pointers 410 that should be enqueued in the application's mbuf_alloc queue 302. The operating system 102 asynchronously replenishes queue 302, making mbufs 402 available to the network application 108. When necessary (e.g., mbuf_alloc 302 is empty), however, applications 108 may use the mbuf_pull primitive 212 to force the system 102 to enqueue a specified (strictly positive) number of mbuf pointers 410 synchronously into mbuf_alloc 302, subject to a timeout interval specified in microseconds (infinite if set to 0). The system 102 asynchronously dequeues pointers 410 from the applications'108 mbuf_dealloc queues 304 and deallocates the respective mbufs. If necessary (e.g., mbuf_dealloc 304 is full), however, an application 108 may use the “mbuf_push” primitive 214, with the tap descriptor argument equal to −1, to force the system 102 to dequeue any pointers 410 from mbuf_dealloc 304 synchronously and deallocate the respective buffers. The primitives mbuf_map 208, mbuf_unmap 210, mbuf_push 212, and mbuf_pull 214 return 0 if successful, or an error code otherwise.
It will be appreciated by those skilled in the art that packets may have fixed or variable length. Although mbufs have a fixed length, a packet of any length can be stored in mbufs. Additionally, a plurality of mbufs can be concatenated to hold a packet that is larger than a single mbuf.
The “interface_tap” primitive 216 allows a NetTap application to open a tap on a specified network interface, as illustrated in
If the interface_tap primitive's 216 “mode” argument is TAP_INPUT, the system 102 enqueues in detour 104A, instead of the regular IP input queue of the system's TCP/IP implementation 112, pointers 410 to packets received by the network interface 114 from the network 116. The system 102 dequeues packet pointers 410 from revert 106A and enqueues them in the regular IP input queue of the system's TCP/IP implementation 112, so that, e.g., a host application 110 may use the system's regular API (e.g., sockets) to receive packets destined to the host, and the system may firewall and forward packets destined to other hosts.
Conversely, if the interface_tap primitive's 216 mode is TAP_OUTPUT, a host application 110 may use the system's regular API to send packets, but the system enqueues in detour 104B packet pointers 410 that TCP/IP implementation 112 would normally pass to the network interface 114 for transmission to the network 116. The system 102 dequeues packet pointers 410 from revert 106B and passes them to the network interface 114 for transmission to the network 116.
Finally, if the interface_tap primitive's 216 mode is TAP_BYPASS, host applications 110 cannot use the system's regular API to send or receive packets via the network interface 114 and the system's firewalling and IP forwarding become inoperative on the network interface 114. The system 102 enqueues in detour 104A pointers 410 to packets received by the network interface 114 from the network 116, and dequeues packet pointers 410 from revert 106B and passes them to the network interface for transmission to the network 116. As shown in
The interface_tap primitive 216 returns a tap descriptor, i.e., a file descriptor (small non-negative integer) that corresponds to the input tap 118 or output tap 120. The “interface_untap” primitive 218 reverts a previous interface_tap 216, and returns 0 if successful, or an error code otherwise. Between corresponding interface_tap 216 and interface_untap 218 calls, the respective tap 118 or 120 is said to be open by the application 108.
Network applications 108 input packets by dequeueing respective pointers 410 from a detour queue 104A or 104B, without system calls. If desired (e.g., detour 104A is empty), however, applications 108 may use the mbuf_pull primitive 212 (with a null number of mbufs) to wait for the system 102 to enqueue a packet pointer 410 in some detour queue 104A or 104B mapped to the application, subject to a specified timeout interval. While the application 108 waits for a packet, the system may run other applications or poll the network interfaces, thereby possibly reducing the number of interrupts.
Conversely, network applications 108 output packets by enqueueing in revert queues 106A or 106B pointers 410 to the mbufs containing the packets. The system 102 asynchronously dequeues mbuf pointers 410 from revert queues 106A and 106B and processes the respective packets (in the TAP_INPUT case, passes them to the IP input queue of the system's TCP/IP implementation 112; in the TAP_OUTPUT and TAP_BYPASS cases, passes them to the network interface 114 for transmission to the network 116). If necessary (e.g., a revert queue 106B is full), however, applications 108 may use the mbuf_push primitive 214 to wait for the system 102 to dequeue a specified number of pointers 410 from a specified revert queue 106A or 106B.
NetTap prevents mbuf leakage as follows. Mbuf headers 404 gain a “pid” field 406 containing the identifier of the process that holds the mbuf 402 (if any). The system 102 updates an mbuf's pid field 406 whenever the system 102 enqueues the mbuf's pointer 410 in an application's mbuf_alloc queue 302 or in a tap's detour queue 104A or 104B. If multiple processes open the same tap, each process must update an mbuf's pid field 406 before dequeueing the mbuf's pointer 410 from detour 104A or 104B. When the system 102 dequeues an mbuf's pointer 410 from an mbuf_dealloc 304 or revert 106A or 106B queue, the system 102 clears the mbuf's pid field 406. Processes gain a flag indicating whether the process has mapped mbufs 402. When a process with such a flag set exits, the system 102 scans the mbuf region to find and deallocate mbufs 402 held by that process.
NetTap uses reference counting for maintaining an interface tap open while at least one process has that tap open. NetTap also keeps a list of taps open by each process. When a process exits, the system automatically untaps any taps that the process still has open.
The processor 502 sends and receives information to and from each of the computer components coupled to the system bus 501 and performs system operations based upon the requirements of the computer system's operating system 102, the NetTap API and network application 108 programs that are installed thereon. The processor 502 may be an INTEL PENTIUM® type microprocessor or the like.
The ROM 510 typically includes a Basic Input-Output system (BIOS) program, which controls basic hardware operations such as the interaction of the microprocessor 502 with the keyboard/mouse of input device 514, the hard disk of storage device 508, or a video display of output devices 51G, and the like. The storage device 508 is a permanent storage medium such as a hard disk, CD-ROM. tape, or the like, which stores the operating system 102, the NetTap API, network application 108 and application program files 518.
The RAM 504 is volatile memory (e.g., SRAM, DRAM and the like). The contents of the RAM 504 may be retrieved from the storage device 508 as required. Illustratively, the RAM 504 is shown with the NetTap API, operating system 102 and network application 108 program concurrently operating.
It will be appreciated by those skilled in the art that although one network application is depicted, more than one network application can be depicted and be within the scope of the invention.
The NetTap API is sent to the RAM 504 for temporary storage and subsequent execution by the processor 502.
The I/O port 512 includes various controllers (not shown) for each input device 514 such as a keyboard, mouse, joystick, and the like, as well as the output devices 516 such as the plurality of network interface 114, adapter, infrared device (not shown) and display 516. Typically, other support circuits 506 include controllers for the storage device 508, floppy drive, graphics display, and the like (not shown).
The operating system 102 may be FreeBSD® or the like. FreeBSD is a stable operating system with mature protocol implementations, state of the art development tools, and a good track record for up to date peripheral support. The operating system is capable of interfacing with all of the hardware components of the computer 500.
In order to oversee the execution of all the files opened, a kernel 524 is stored in the RAM 504. The kernel 524 is installed at dedicated addresses in the RAM 504 and is responsible for memory management, process and task management, and disk management. The kernel 524 manages one or more files that are open during the course of operation.
The RAM 504 further comprises a portion of memory for storing detour queue queues 104A, 104B and revert queues 106A and 106B and mbuf 402 which comprises mbuf_alloc 302, mbuf dealloc 304.
The processor 602 may be, e.g., an Intel PENTIUM® microprocessor, and executes instructions from the operating system 102 and application 108 and programs 518. The operating system may be, e.g., FreeBSD, and incorporates the NetTap API. Network applications 108 use the NetTap API to send or receive packets. Programs and data are stored in main memory 604 while being processed. In particular, mbufs 402, mbuf_alloc queues 302, mbuf_dealloc queues 304, detour queues 104A, 104B, and revert queues 106A, 106B are stored in main memory 604. The processor 602 typically copies programs and initial data from hard disk 620 to main memory 604 during system or application initialization. The video 608, keyboard 610, and mouse 612 provide system interaction with a user, e.g. a system administrator. Network applications 108 must be executed by a system administrator. The network interface cards 520 and 522 transfer packets between the respective networks and mbufs in main memory 604. The network interface cards 520 and 522 may be controlled by processor 602 and need not have a coprocessor. The number of network interface cards in a system may vary from the two shown in
It will be appreciated by those skilled in the art that NetTap applications can tap any number of network interfaces and therefore can implement host, bridge or router configurations. Moreover, NetTap applications 108 can be written in any language and implement arbitrary functionality.
Router or host configurations may need additional protocol support, provided, e.g., by user-level libraries for TCP/IP protocols, IP security, or output link scheduling. Such libraries can run on top of the NetTap API. Applications may easily specialize or modify user-level libraries. For example, applications such as NAT (network address translation) or LSNAT (load sharing using IP network address translation) might modify a user-level IP implementation.
NetTap applications 108 may run with real-time priority to obtain guaranteed performance regardless of other load (e.g., host applications) on the system. Real-time priorities are always greater than time-sharing priorities. FreeBSD has supported real-time priorities since version 3.0. NetTap applications 108 that need multiple threads of control may use POSIX (user-level) threads on top of a single real-time process, so as to avoid context switch overheads. Single-thread event-driven implementations can be expected to outperform multiple-thread implementations, however.
On the other hand, if a single processor 602 is unable to keep up with network traffic tapped by an application 108, it may be useful to run multiple instances of the application 108 on a shared-memory multiprocessor system, with one application instance per processor. Multiple applications 108 can map mbufs 402 and tap each one of the same plurality of network interface 114. In addition, mbuf_alloc 302 and mbuf_dealloc 304 queues are private to each application 108 currently being run. However, detour 104A and 104 b and revert 106A and 106B queues are shared by all applications 108 that open the respective taps 118 or 120. Concurrent accesses to tapping queues or other shared data structures must be synchronized. Because NetTap is PC-based, synchronization can be achieved at user level, without system call overheads. The concurrent applications 108 can define a shared integer to be a lock that guards certain shared data structures. Each application 108 uses, for example, “i486” (or later) CPU's CMPXCHG (compare and exchange) instruction to acquire or release the lock respectively before or after accessing the shared data structures. The system 102 keeps a list of locks used by each application 108. When the application 108 exits, the system 102 automatically releases any locks still held by the application.
While the foregoing description represents preferred embodiments, it will be obvious to those skilled in the art that various changes and modifications may be made, without departing from the spirit and scope of the invention pointed out by the following claims:
|Brevet cité||Date de dépôt||Date de publication||Déposant||Titre|
|US5491800 *||20 déc. 1993||13 févr. 1996||Taligent, Inc.||Object-oriented remote procedure call networking system|
|US6064805 *||2 juil. 1997||16 mai 2000||Unisys Corporation||Method, system, and computer program product for intraconnect data communication using buffer pools and buffer pool management|
|US6145061 *||7 janv. 1998||7 nov. 2000||Tandem Computers Incorporated||Method of management of a circular queue for asynchronous access|
|US6282581 *||27 mars 1997||28 août 2001||Hewlett-Packard Company||Mechanism for resource allocation and for dispatching incoming calls in a distributed object environment|
|US6546413 *||14 déc. 1998||8 avr. 2003||Charles J. Northrup||Access-method-independent exchange using a communication primitive|
|US6675218 *||14 août 1998||6 janv. 2004||3Com Corporation||System for user-space network packet modification|
|1||Blott et al., "NetTap: An Efficient and Reliable PC-Based Platform for Network Programming," published Mar. 26, 2000 in the Institute of Electrical and Electronic Engineers Proceedings of OPENARCH '2000.|
|2||Blumrich et al., Virtual Memory Mapped Network Interface for the SHRIMP Multicomputer, In Proceedings of the 21st Annual International Symposium on Computer Architecture, Apr. 1994, pp. 142-153.|
|3||Druschel et al., Experiences with a High-Speed Network Adaptor: A Software Perspective. In Proc. of SIGCOMM-94, pp. 2-13, Aug. 1994.|
|4||G. Lehey, The Complete FreeBSD, Walnut Creek CD ROM Books (2d ed. 1997).|
|5||McKusick et al., the Design and Implementation of the 4.4 BSD Operating System, Addison-Wesley (1996).|
|6||von Eicken et al., U-Net: A User-Level Network Interface for Parallel and Distributed Computing, In SIGOPS '95, pp. 40-53, Dec. 1995 CO., USA.|
|7||Welsh et al., ATM and Fast Ethernet Network Interfaces for User-level Communication. 1997 IEEE, Published in Proceedings of the Third International Symposium on High Performance Computer Architecture, Feb. 1-5, 1997 in San Antonio, Texas, USA.|
|Brevet citant||Date de dépôt||Date de publication||Déposant||Titre|
|US7571464 *||27 août 2004||4 août 2009||International Business Machines Corporation||Secure bidirectional cross-system communications framework|
|US7779411 *||17 févr. 2005||17 août 2010||Red Hat, Inc.||System, method and medium for providing asynchronous input and output with less system calls to and from an operating system|
|US7861253 *||18 nov. 2004||28 déc. 2010||Microstrategy, Inc.||Systems and methods for accessing a business intelligence system through a business productivity client|
|US7911994 *||17 déc. 2003||22 mars 2011||Openwave Systems Inc.||Confirmation of delivery of content to an HTTP/TCP device|
|US8028302 *||29 oct. 2007||27 sept. 2011||Whirlpool Corporation||Software architecture system with embedded virtual router|
|US8141077||12 mars 2010||20 mars 2012||Red Hat, Inc.||System, method and medium for providing asynchronous input and output with less system calls to and from an operating system|
|US8542582||23 oct. 2009||24 sept. 2013||Unwired Planet, Llc||Confirmation of delivery of content to an HTTP/TCP device|
|US8752063||23 juin 2011||10 juin 2014||Microsoft Corporation||Programming interface for data communications|
|US20040205231 *||17 déc. 2003||14 oct. 2004||James Clarke||Confirmation of delivery of content to an HTTP/TCP device|
|US20060048217 *||27 août 2004||2 mars 2006||International Business Machines Corporation||Secure bidirectional cross-system communications framework|
|US20060184948 *||17 févr. 2005||17 août 2006||Red Hat, Inc.||System, method and medium for providing asynchronous input and output with less system calls to and from an operating system|
|US20070067488 *||16 sept. 2005||22 mars 2007||Ebay Inc.||System and method for transferring data|
|US20080109830 *||29 oct. 2007||8 mai 2008||Whirlpool Corporation||Software Architecture System with Embedded Virtual Router|
|US20080126956 *||4 août 2006||29 mai 2008||Kodosky Jeffrey L||Asynchronous Wires for Graphical Programming|
|US20090249371 *||28 mars 2008||1 oct. 2009||Omar Cardona||Buffer allocation for network subsystem|
|US20100042739 *||18 févr. 2010||Openwave Systems Inc.||Confirmation of delivery of content to an http/tcp device|
|US20100153974 *||16 déc. 2008||17 juin 2010||International Business Machines Corporation||Obtain buffers for an input/output driver|
|US20100169897 *||12 mars 2010||1 juil. 2010||Red Hat, Inc.||System, Method and Medium for Providing Asynchronous Input and Output with Less System Calls to and From an Operating System|
|Classification aux États-Unis||709/238, 719/328, 719/314, 709/244|
|Classification internationale||H04L29/06, H04L12/56, H04L29/08, G06F15/173|
|Classification coopérative||H04L69/16, H04L69/162, H04L69/329, H04L49/90, H04L49/901, H04L29/06|
|Classification européenne||H04L29/06J3S, H04L49/90C, H04L49/90, H04L29/06J, H04L29/06|
|25 mai 2001||AS||Assignment|
Owner name: LUCENT TECHNOLOGIES INC., NEW JERSEY
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BLOTT, STEPHEN M.;BRUSTOLONI, JOSE CARLOS;MARTIN, CLIFFORD E.;REEL/FRAME:012127/0839;SIGNING DATES FROM 20010518 TO 20010522
|11 nov. 2010||FPAY||Fee payment|
Year of fee payment: 4
|6 nov. 2014||FPAY||Fee payment|
Year of fee payment: 8